Most of my deployments in other cloud providers are usually just 2-3 servers which can probably all be put into a single Akash deploy.yml, probably without much technical consideration. But I do have multi-hundred server deployments so I can lend a few items to the discussion based on experience.
Some cases which could use “extra options” might be:
- Specifying which servers should NOT be placed on the same hardware, yet in the same physical connection area (AWS AZ idea)
- Specifying which servers should be on the same server (probably not urgent), but perhaps a “loose” preference would be enough
- Specifying back-end networks for different enclaves/tiers of private communication (vxlan, backend communication between database nodes that do not need to be public)
And for something extra fun, typically large scale deployments don’t happen all at once, but take months to build.
It would be good to be able to deploy additional servers into the same vxlan/vpc/private network as other existing containers. Maybe not so easy to do, but perhaps adding some guidance on the site as to a recommended strategy for modifying a deployment would remove this as a tripping point for new users.
For placement, the example is “westcoast” but the documentation does not clearly define whether that is the only choice for now, or one of many available choices. I would have thought that the superminis might be “all over” in which case it would be cool to pick “global/us/asia/europe-supermini” with at least “bandwidth: 50mbps” for example.
Obviously these are no small tasks, but it is equally as helpful to document “This is on the roadmap, don’t try it now”.
This way new onboarding folks don’t waste time trying to figure out how to make something impossible work.
Instead they can be guided to use the best functioning parts of the system first.
(BTW looks like it’s a “known thing” with various paths to the goal Akash Development Roadmap · GitHub)